D-4243-3

Teaching System Dynamics:

Looking at Epidemics

William Glass-Husain

System Dynamics in Education Project

Massachusetts Institute of Technology
System Dynamics Group, E40-294
1 Amherst St.
Cambridge, MA 02139

William Glass, 1991

October 3, 1991
with slight revisions February 12, 1992
and July 8, 1993

D-4243-3

The author would like to thank the following members of the System Dynamics in
Education Project for their comments, criticisms and suggestions.
Rebecca Niles
Dave Wooley
Chris Prince

In addition, the author is deeply indebted to Dr. Jay Forrester and Dr. Pl Davidson, for
their ideas, guidance, and inspiration.

D-4243-3

Table of Contents

Introduction: Epidemics, System Dynamics, and the Learning Process

Lesson 1:

The Epidemic Game:

A)
B)

Lesson 2:

Teacher notes and ideas

Handouts, transparencies for class

Modelling an Epidemic
A) Teacher notes: positive and negative feedback
B) Worksheet: Modelling an epidemic

D-4243-3

D-4243-3

Introduction:

Epidemics, System Dynamics and

the Learning Process

Why Study Epidemics?

The lessons presented here use the tools of system dynamics to explore epidemics. When we refer to an
epidemic we are referring not just to the number of people infected with a disease, but also to the dynamics
of the epidemic; in other words, to how the number changes through time. Epidemics can involve any one
of a number of contagious diseases, or can even involve the infection of some other item that spreads
from one person to another -- such as an idea or information that spreads through word of mouth.
The reason for teaching the dynamics of epidemics is not because of a sudden decline in the number
of working epidemiologists, nor is the sole goal of this series of exercises to teach students how to build
good computer models.

Instead, the intent of these papers, and of the curriculum they describe, is to use

system dynamics as a tool to help people learn to think.

Thinking may seem like a natural process no one needs no learn; in particular, this is true when
the term is used to mean the process of being conscious. However, thinking as a process of
conceptualization and reflection is discouraged by traditional fact-based, teacher to student education that
actively discourages the continual questioning, exploration, and discovery that is needed to gain
understanding and appreciation of how the world works. Educators need to encourage innovation and
develop new methods of instruction to prepare students to face the diversity of today's changing world.

Goals for the Learning Process

To fulfill the rather broad challenge implied in the preceding paragraph, the author suggests four goals that
should be incorporated into the learning process.
Spiral instruction -- No concept can be learned from one source or in one sitting. The best
method of learning is to see the concept presented in many different forms in many
different contexts. The ideal process of learning is rather like a spiral, with students
exploring different areas in different ways, while connections are continually drawn to
central themes by the teacher, the student or (ideally) both. The term spiral comes
from the continual circling back and reinforcing of old concepts as new ones are
introduced. This exploration can be done by discussion, by reading, by writing, and by
hands-on experimentation and simulation.
Learner-directed learning -- Something that should go hand in hand with this spiral method of study
is learner-directed learning. The teacher should not be solely responsible for directing
every loop of the spiral.

Discovery of new ideas and new connections to old themes is

what makes any subject interesting and exciting and hence more likely to be understood
and appreciated. Closely related to this is the idea of inter-activeness -- that students get a

D-4243-3

chance to interact with each other and learn (at least some of the time) in a hands-on
fashion.
Relate the concrete to the theoretical -- These two perspectives are complements of each other, not
opposites, and should both be included. Learning should be done in context. Examples
used should be connected to a framework already held by the student. People learn new
things best when they are taught in relation to other things they are already familiar with.
However, theory is important. Through the learning of theoretical principles a more
general understanding can be gained.
Scientific method -- Finally, a key concept in learning is that of the scientific method. Before any
kind of experimentation, hypotheses should be made of the behavior of the experiment.
After the experiment, there should be a follow up evaluation asking the question What
happened? and examining how the behavior of this experiment relates to previously
learned concepts and ideas. This is particularly important when using a computer to do
simulation. Without this process of introspection, students may fall prey to the video
game syndrome of pushing buttons to win the simulation and may not fully stop to
consider the larger issues involved.

Studying an Epidemic with System Dynamics

Using system dynamics to explore epidemics, all four of these goals can be approached.
In these lessons many general characteristics of epidemics will be revealed. Yet. the lessons also
communicate on a deeper level several fundamental truths. Concepts such as stock/flow models, feedback
loops, shifts in dominance, system boundaries, and aggregation will be discovered as the system is explored
on many different levels. Other ideas such as the relationship between structure and behavior, the validity
of different paradigms for understanding different aspects of a system, and the endogenous nature of behavior
are also subtly but strongly interwoven themes that will be touched upon over and over again during the
examination of this simple-seeming system.
Different modes of teaching can also be employed. These range from discussion, to presentation of
real-world data, to guided and independent modeling and simulation.
Learner-directed learning should be a key component of teaching system dynamics. It is
particularly suited to the individual nature of modeling and simulation. When students construct their own
models, they can learn at their own pace (within broad guidelines), and will be able to develop their own
policies to simulate. Also, because the very nature of system dynamics demands that it cannot be taught as
a collection of unrelated facts, students are actively encouraged to place their pieces of information they
learn into a framework and apply it while modeling and thinking in general.
All simulation of dynamic behavior should be preceded by students making sketches of what they
predict the behavior to be. There should be discussion as to the reasons for their predictions. In the case of

D-4243-3

disagreement, it is often advantageous to set one side against the other and let the students argue (in a
guided, controlled fashion). The author has found this technique particularly effective when the correct
answer is some combination of both answers proposed (i.e. both sides are right). Each group will try to
convince the other, and eventually there will usually be a consensus as to what the behavior should be. It
is important to guide the argument so that all students have a chance to present their views, and it is good
to clarify points when necessary.

Structure of Lessons
While the sequence can stop at any point, the each of the lessons presented in this packet are designed to
build on concepts introduced in the previous lessons. Students participating in lesson 1, the Epidemic
Game, will have a valuable learning experience, but those continuing on to lessons 2 and 3 will have a
deeper understanding and appreciation of how both an epidemic and system dynamics work.
Lesson 1 : The Epidemic Game (time: one hour) -- This lesson approaches both system
dynamics and an epidemic from an intuitive perspective. Students play a short simulation
game, then engage in discussion to figure out how the epidemic works.

Students are

encouraged to be critical thinkers, creating and contrasting different hypothesizes until

they cooperatively discover an explanation that makes sense. Students discover that
different processes cause different behaviors -- a notion fundamental to all systems -- and
that two different processes operate in an epidemic causing different behaviors at different
times.
Lesson 2 : Modeling an Epidemic (time: four hours) -- This lesson has two parts. The
first is a short lecture and discussion on the basic principles of feedback in systems.
Several examples are explored. In the second part, students follow a worksheet and,
working in pairs on Macintoshes, build a model of a generic epidemic. Through doing
this they apply many of the general concepts discussed in class to a specific example.
Lesson 3: Epidemic Research Project (time: variable) -- A suggested final lesson is an
optional project in which students do research on other, historical epidemics. The
students then create more sophisticated, accurate epidemic models based on their research.
The lesson could end with a written report and oral presentation of the completed model.
(not included in packet)

D-4243-3

Lesson 1: Playing the Epidemic Game

Why Play the Game?
The epidemic game is a role playing simulation in which the students one by
one become infected with a fictitious malady.

The game and debriefing for

the game can be used independently in the classroom to give an intuitive

introduction to how an epidemic works, or it can be combined with the next
lesson, Modeling an Epidemic, that uses a slightly more formal, complementary
approach of building a computer model of an epidemic.
The epidemic game will give students an intuitive, hands-on
introduction to the following concepts:
1) Dynamics of an epidemic
2) Influence of structure -- rules of the game -- on systems
behavior
3) Feedback in dynamic systems

The Epidemic Game

The epidemic game is outlined in a set of rules given in Appendix II.

While

the instructions may seem complicated, the idea is simple enough. One student
begins the game infected with a disease.

While interacting through

handshakes with the rest of the group that student begins the epidemic that
results in everyone becoming infected. The time at which each person
becomes infected is privately recorded. Later, this information will be
collected and graphed to show the behavior of the system. After the game is
played, students are challenged to determine the dynamics of the game, in
other words, the changes in the cumulative number of infections during the
course of the game. Then, using the recorded data, the number of new
infections for each turn is tallied up to produce graphs of both the actual
number of new infections and the number of cumulative infections versus
time. Finally, a computer is brought in with a model of the epidemic that can
simulate almost identical behavior to the game just played.
A typical session with the epidemic game will go in these steps:

D-4243-3

Read the rules. Skip the description of what the disease actually is. Tell
them the game will run for twenty turns, each turn representing
one month.
Choose an initial infected person. Have all the students close their eyes
and tap one student on the shoulder to indicate that he or she is
the first infected student.
Play the game.

Count off turns, writing the month number on the

board each turn, and watch each student shakes hands twice (or
to be more precise, each hand once). Then, sometime around the
twelfth turn when everyone should be infected, stop the game.
Begin the discussion.
There are a couple of features of the game that need a bit more
elaboration. First of all, because of the limited size of many classes, each
student represents two people. An easy way for people to keep track of both
identities is to picture them as their two hands. Each hand should shake hands
with one other hand every term (either the left or right hand can shake any
other hand). If a healthy person shakes hands with a Sick Person and
receives the secret handshake, that hand will become infected. It is
important to emphasize when you read the rules that students should keep
track of the month in the game when each hand becomes infected, as much of
the rest of the time after the game is over will be spent discussing that data.
A second slightly tricky aspect of the game is that, just like the real
world, not every contact between infected people and non-infected people
brings a new infection. In the epidemic game, when a Sick Person shakes
hands with a Healthy Person, there is only a 50% probability that the Healthy
Person will be infected.

To ensure this, before each turn begins, each Sick

Person picks either the number 1 or 2 for each infected hand. The leader of
the game then picks the 1 or 2 and announced the number, as he or she writes
the month number on the board. For each hand, if the Sick Persons number
matches the leaders number, the Sick Person gives the secret handshake. The
probability can be varied by changing the possible range of numbers both the
leader and the Sick Person choose from.
Finally, because it is important that the rules be followed exactly in
order for the game to produce the correct results, it is a good idea to stop
several turns into the game, tell everyone to close their eyes, and then have

10

D-4243-3

the Sick People raise their hands. If their are more than two or three Sick
People at this point, something is wrong. Also, do a similar head count around
the tenth or eleventh turn to determine if it is time to halt the game or not.

Discussion: What Happened? Why Did It Happen?

The discussion that follows the actual simulation is probably the most
important part of the exercise. During the discussion, students need to take
their raw experiences from the game and put it into a systems framework -enabling them to understand how and why this epidemic worked.
As a teacher, your role is important, but difficult. While the temptation
may be great to tell your students what to see, they will understand these
concepts best if they figure them out on their own under your guidance. Your
role for the purposes of this discussion will be as a facilitator than a lecturer.
Facilitating the discussion will involve basic issues such as making sure
everyone gets a chance to speak. Clarification and restatement of what
students are saying is often necessary. At times, however, you may just need to
stand to the side and let students talk it out without your explicit involvement.
You should, however, guide the discussion through several steps:
Step 1: Making a first hypothesis
Step 2: Presentation of hypotheses
Step 3: Reason for hypotheses
Step 4: Resolution
Step 5: Graph of results from game play
Additionally, there are a few general issues involving system thinking
that can run through the discussion as well. Finally, if students do not
mention some of the issues you would like to see discussed, or they seem stuck
in a particular rut of thinking, you will probably want to get more involved..
Step 1: Making a first hypothesis -- To begin the discussion, ask your
students to draw their impression of how the number of
infections changed through time. Draw a sample axis on the
blackboard, label the y-axis total infections and the x-axis
time and ask the students to fill out a similar axis on paper by
themselves. When they are done ask them to write a sentence or
two of why they think their guess is correct.

D-4243-3

11

Step 2: Presentation of hypotheses -- Have some students put their

hypotheses on the black board. You might want to select the ones
for presentation based on uniqueness, e.g. if 5 people all guess
exponential growth there is no need for more than one to be
drawn on the board. You should have guesses split between
Figure 1 (exponential growth) and Figure 2 (goal-seeking
growth), with a few people guessing figure 3 (linear growth) and
Figure 4 (s-shaped growth). If for some reason, no students
predict behavior similar to either figure 1 or 2 you may want to
draw them and present arguments in favor of them yourself.
Step 3: Reason for hypotheses -- Now, go back through the students who
graphed their predictions on the board and ask each one to give a
short argument in favor of their theory.
Step 4: Resolution -- Guide the discussion so that students realize that
both the arguments for Figure 1 and figure two have merit in
some circumstances, and not in others, and that in fact the
answer is Figure 4, a combination of figures 1 and 2.
Step 5: Graph Results from Game Play -- Create a graph based upon the
recorded data of the number of new infections each turn
(infection rate). Add up the new infections and create a graph
of the total number of infections so far for each turn
(cumulative infections). This can either be done by hand on
the blackboard, or using a spreadsheet program on a Macintosh
such as Excel. Compare the actual behavior with the behavior
predicted by the classs hypothesis.
In this discussion students get their first taste of several important system
concepts:
circular feedback nature of systems
distinction between flow variables that indicate
change from one time period to another (the number
of new infections) and stock variables that
accumulate (number of total infections)
the use of different models to explain different
aspects of a system

12

D-4243-3

the need to explain the behavior of systems (the

dynamics of the game) in terms of the structural
interrelationships between the elements of the
system (the rules of the game)
Each of these are themes that can be presented in many different systems and
at many different levels but that are particularly explicit in the epidemic
game.

Figure 1: Exponential Growth

Figure 2: Goal Seeking Growth

Figure 3: Linear Growth

Figure 4: S-shaped Growth

Epidemic Behavior: The Right Answer

Figures 1 through 4 show possible graphs of how the stock of Sick People
varied through the game. Out of the four possibilities in Figures 1 through 4,
which is the right answer students should decide on?

D-4243-3

13

We can eliminate linear growth, Figure 3, pretty quickly. As the

number of Sick People and Healthy People varies, the number of infecting
handshakes per turn, and hence the new number of nerds, is going to change.
This means that the growth in the Sick People population cannot be constant,
in other words, the graph cannot be a straight diagonal line. Lets study some
other possibilities by examining Figures 1 and 2 in more detail.
Exponential growth has some appealing evidence in favor of it.
Remember, infected people cause others to be infected. The more infected
people, the higher the infection rate (number of new people infected each
turn). This means that the infection rate has got to be very low when there
are only a few people already infected.

As the number of infected people

rises, more infectious handshakes will be given, and the infection rate will
rise.
Sounds good, right? Unfortunately, there is a serious problem here.
What happens when you run out of people?

Students arguing in favor of

goal-seeking growth (figure 2) will be quick to point out that after they
caught the disease, they received many infecting handshakes that had no
effect because they had already caught the disease. In fact, by the end of the
simulation very few people are left who have not yet been infected, and most
infecting handshakes delivered are useless. This means that the number of
new people being infected must slow down as the number of Sick People rises.
While the number of infected people never goes down (there are no cures in
this simulation), as the un-infected population decreases, the infection rate
decreases.
SICK PEOPLE

HEALTHY POPULATION

(infected population)

New SICK PEOPLE each

turn ---- (infection rate)

New SICK PEOPLE each

turn ---- (infection rate)

(a)
(b)
Figure 5: (a) Argument for exponential growth: As the number of Sick People
increase, the infection rate increases; (b) Argument for goal-seeking

14

D-4243-3

behavior: As the number of Healthy People decrease (and the Sick People
increase), the infection rate decreases.*
This justification for goal-seeking behavior also has a serious flaw.
According to the graph in Figure 2, new infections are large in the first turn
but slowly drops to zero by the time that there are no un-infected people left.
But, we know that there can only be a maximum of one infection the first turn
-- after all, the game begins with only one Sick Person to do the infecting.
Both exponential growth and goal-seeking behavior seem to fit
different aspects of our game, but each has flaws. The exponential growth
does not adequately explain the behavior near the end of the simulation, while
the un-infected population is low. The goal-seeking growth does not seem to
fit in the beginning when the infected population is low.
The answer is that both Figure 1 and 2 are correct answers, but for
different times in the simulation. Each is a different aspect of the full
behavior of S-shaped growth, shown in Figure 4. Early on in the simulation,
when the upper limit is very far away, Figure 5a is the most prominent rule
that determines the behavior, and we see near-exponential growth. However,
as the infected population grows nearer to the upper limit, the explanation
shown in figure 5b becomes important, and the growth slows down and
becomes goal-seeking.

* It is important to note that the relationship between the SICK PEOPLE or the HEALTHY PEOPLE and

the infection rate is circular and not just one way. For example, if the number of SICK PEOPLE increase,
the infection rate will increase. This then causes an even greater increase in the number of SICK PEOPLE.
This idea of feedback will be examined in great detail in the next chapter.

D-4243-3

15

number of Healthy People

relative to total population

"Importance" of Sick People infecting

"Importance" of upper limit

SICK PEOPLE
(infected population)

HEALTHY PEOPLE
(infected population)

New SICK PEOPLE

each turn
(infection rate)

New SICK PEOPLE

each turn
(infection rate)

Figure 6: As the Sick population rises, the influence of the number of Sick
People on the infection rate becomes weaker and the influence of the number
of Healthy People on the infection rate becomes stronger.

most Sick People

infecting Healthy
People

Infected Population (people)

32

most Sick People interacting

with other Sick People

24
16

8
0

6
Time (months)

12

Figure 7: The S-shaped growth caused by the shift from exponential growth to
goal-seeking behavior

The Final Reinforcement

16

D-4243-3

An important ending to this exercise is to let students experiment with a

STELLA model of an epidemic such as the one presented in Appendix III.
Through guided experimentation, students will gain a fuller appreciation of
the behavior produced by the epidemic system.
When presenting the model, give a quick explanation of how the model
works and then show students how to vary the parameters in the model, to
simulate the model, and to graph the simulation. The parameters to vary are
the initial number of Sick People and Healthy People, the probability of
infection per contact between a Sick Person and a Sick Person, and the
number of handshakes per person per turn. Students should discover that
except for trivial cases such as an initial number of zero Sick People,
changing the inputs of the epidemic has no effect upon the behavior pattern
of the epidemic (although sometimes the scale or length of the simulation will
need to be adjusted to show full S-shaped growth).

D-4243-3

Appendix I: Sample Outline of Epidemic Lesson 1

I. Play Game
II. Guess Dynamics
A. Individually
B. Arrive at Consensus
III. Graph Actual Dynamics of Game
IV. Experiment with Computer Model

17

18

D-4243-3

Appendix II: The Epidemic Game

Background
In this game we are going to simulate an epidemic of a disease. One person
among you has just been infected. Most of you may think that one person is no
big deal, but medical experts versed in system dynamics are concerned, and
have begun a program of research. They have determined that this disease is
an unusual disease for three reasons. First, the disease is only spread through
the shaking of hands. Second, the symptoms do not appear for three years
after infection even though an infected person is contagious immediately.
Third, infections are not guaranteed to spread with every contact of the
disease; rather there is a probability of infection with every contact.
The symptoms of the disease are quite striking. The first signs of the
disease is a slight slouching, usually beginning around 3 years after infection.
After that the bone structure of the hips is altered, most noticeably resulting
in a tendency for one to pull one's pants as high as possible. Other features
include a smoothing of the nose-skin (which can be corrected by placing
adhesive at strategic points on eye wear) and craving for tee-shirts with
obscure mathematical symbols. This sounds minor, but usually culminates in
attacks so severe victims must be confined to the famous treatment center, the
Masochistic Institute of Terminology.

Rules of the Game

In order to increase the playing population in the simulation, think of
each of your hands as a separate person and repeat the steps of the game for
each hand. With minutes simulating months, the game will be run at one
minute intervals allowing one contact per person per minute with a 50%
infection probability. The disease will be passed by using the secret
handshake, which your teacher will demonstrate.
Each Time Period:
1)

Designate your left hand or right hand as the first to shake during the
period.

2) Select a random person in the room with whom to shake hands.

3) If your designated hand is infected, pick either the number 1 or 2 and
remember your selection.
4) If your number is called and your designated hand is infected, prepare to
administer the secret handshake on your next turn.
5) Shake the hand of a random person.
6) Record on your score card the infection status of your designated hand.
7) Repeat steps 2 through 6 using your non-designated hand.
8) Go back to step 1 and repeat process for next time period.

D-4243-3

19

Appendix III: Sample Epidemic Model

Healthy People

Prob of a Sick Person Meeting a Healthy

Sick people

New Infections

Handshakes Per Round

Seconds Per Handshake

Seconds Per Round

Healthy_People(t) = Healthy_People(t - dt) + (- New_Infections) * dt

INIT Healthy_People = 26 {People}
New_Infections =
Prob_of_a_Sick_Person_Meeting_a_Healthy*Handshakes_Per_Round*Sic
k_people {Infections/round}
Sick_people(t) = Sick_people(t - dt) + (New_Infections) * dt
INIT Sick_people = 1 {People}
New_Infections =
Prob_of_a_Sick_Person_Meeting_a_Healthy*Handshakes_Per_Round*Sic
k_people {Infections/round}
Handshakes_Per_Round = Seconds_Per_Round/Seconds_Per_Handshake
{(handshakes/round)/person}
Prob_of_a_Sick_Person_Meeting_a_Healthy =
Healthy_People/(Healthy_People+Sick_people) {dimensionless}
Seconds_Per_Handshake = 5 {seconds/handshake}
Seconds_Per_Round = 10 {seconds/round}

20

D-4243-3

Epidemic lesson 1 / transparency 1

Rules of the Game

Remember:

Each of your hands is a separate person.

+ So, repeat the steps of the game for each hand.
Each minute represents one month.
For each Sick Person handshake, 50% probability of
infection.
Sick People use secret handshake when infecting.
Each Time Period
1) Designate your left hand or right hand as the first to shake
during the period.
2) Select a random person in the room with whom to shake

hands.
3) If your designated hand is infected, pick either the number 1 or 2
and remember your selection.
4) If your number is called and your designated hand is infected,
prepare to administer the secret handshake on your next turn.
5) Shake the hand of a random person.
6) Record on your score card the infection status of your designated
hand.
7) Repeat steps 2 through 6 using your non-designated hand.
8) Go back to step 1 and repeat process for next time period.

D-4243-3

21

Epidemic lesson 1 / handout 1

Name:________________
Game

Epidemic

Score Card

(fill out during the game)

Month
Right

Left

Right

Month

1
2
3
4
5
6
7
8
9
10

Left

11
12
13
14
15
16
17
18
19
20

**********************************************************************

Hypothesized System Behavior

(fill out after game)

On the axis below, sketch how you think the percentage of people in the
infected population changed over time.

Percent of Population which is Infected

100
80
60
40
20
3

12

15

18

Time
Write a sentence or two describing why you believe the epidemic
showed the behavior sketched above.

22
Epidemic lesson 1 / transparency 2

D-4243-3

D-4243-3

23

Lesson 2: Modeling an Epidemic

Goals of Lesson:
In the last lesson, students engaged in an intuitive exploration of the structure of the epidemic the set
of interrelations that make up the epidemic system. In this lesson, students gain understanding of the
connection between the structure and the dynamics of an epidemic by modeling the epidemic from the
previous lesson.

At the same time as they learn about the epidemic students will learn these system

skills:
1) Ability to discern feedback loops in different situations,
2) Ability to explain behavior of a one or two loop system in terms of structure,
3) Basic familiarity with STELLA building blocks, and
4) Experience building in stages a small model using STELLA.

Format of Lesson
This lesson is organized in two parts. The first half involves a short introduction to the terminology and
concepts of feedback and STELLA modeling, presented by the teacher. During the second half, students
work on Macintosh computers in pairs and build a model of the epidemic by following a highly structured,
hands-on worksheet. By building the model, students gain self-confidence and motivation as they form a
deeper understanding of how the epidemic works.

Feedback loops
Begin the class by pointing out the structural importance of feedback in driving the behavior of the
epidemic simulated in the previous lesson. In the simulation, Sick People infected Healthy People
creating more Sick People. Just as important, however, those newly infected Sick People helped to create
even more Sick People! Similarly, there was another kind of feedback in the later stages of an epidemic.
More Healthy People being infected meant less Healthy People stayed around, which meant less Sick
People became infected. The circular aspect of each structure was a key factor in producing in both the
early exponential growth in infections and the infection rate and the leveling off of infections near the end.

Positive feedback
Next, explain the general concept of both positive and negative feedback. Positive feedback is a selfreinforcing process that causes exponential growth. A good first example of positive feedback is something
that should be familiar to most of your students: a savings account (see Figure 1). Many people view a
bank account as a straight cause and effect system. Savings generate interest. However, a bank account is

24

D-4243-3

actually a positive feedback loop: savings in a bank generates interest, which gets added to the savings,
generating more interest.
Other good examples of positive feedback include a population of rabbits and a shouting match.
Rabbits give birth to produce more rabbits. These new rabbits give birth to even more rabbits. One person
shouts at another, causing the other person to become angry and shout back. This results in the original
person becoming angrier and shouting even more. Each of these systems is often thought of as a one way
relationship but is really a positive loop.

Savings

Anger

+
Interest

Rabbits

+
Shouting

+
Births

Figure 1: Three cause and effect open loops.

We can represent these systems with causal loop diagrams (Figure 2). The + sign next to the
arrows indicates that as the first item changes, the second changes in the same direction. Thus, more
savings will cause more interest, more anger means more shouting, and more rabbits will produce more
births. Similarly, less savings causes less interest, less anger causes less shouting, and fewer rabbits cause
fewer births. The + sign within the circular arrows in the feedback loops tells us that these are positive
feedback loops.

Savings

Anger

+
+

Interest
Figure 2: Positive feedback loops

Rabbits

+
+

+
Shouting

Births

D-4243-3

25

Figure 3: Savings in a bank with 6% interest

If we look at the dynamics of a variable in a positive feedback loop, we see the behavior pattern of
exponential growth. For example, in Figure 3 is a graph of a savings account that initially had $20
deposited, with an interest rate of 6%. In 60 years, the amount of savings has increased to $660, 33 times
the original amount!
This exponential growth occurs in the savings/interest loop of the positive feedback structure.
Each time unit, the interest earned causes a increase in the savings of a certain increment. As the interest
is proportional to the amount of savings, in the next time unit there will be a greater amount of interest,
causing an ever greater increment in the amount of savings. The positive feedback causes both the savings
and the rate that savings increases (the interest) to increase exponentially. The same behavior occurs in the
cases of a shouting match and rabbit populations as seen in Figure 2.

Negative Feedback
Unlike self-reinforcing positive feedback, negative feedback is a process that adjusts a variable to a goal. In
various disciplines, this can be referred to as a homeostatic, self-governing, or goal-seeking process.
In Figure 4 are several causal diagrams of systems containing negative feedback. Note that here
the sign by the arrows refers to changes in the opposite direction. For example, in a house with
central heating, if the temperature of the house drops the gap between the temperature and the desired
temperature will increase. As the gap increases, the furnace will increase its heat production. As the
furnace produces more heat, the temperature will begin to rise again and the gap will down, lowering the
heat production until the gap is zero and the furnace shuts off.

26

D-4243-3

(a)
(b)
Figure 4: (a) central heating system (b) insulin regulation of blood sugar
Similarly, if a person increases the amount of sugar in their blood, by eating a candy bar, the gap
between the desired concentration of sugar in the blood and the actual concentration will increase. The
increased gap will cause the production of insulin by the pancreas. Insulin will then increase the rate of
sugar absorption by body cells from the blood stream, lowering the amount of sugar in the blood.
Both the central heating system and the insulin/glucose system show goal-seeking behavior. In
each case there is a desired goal that the system gradually approaches. The system changes most when it is
far from the goal and least when it is close to it. In the central heating system the temperature is initially
lower than the desired temperature, and rises to meet it. In the insulin/glucose system, the blood sugar is
higher than desired, and is lowered until it is at the correct level.
Something that is important to remember is that the "+" sign does not necessarily mean that one
variable causes another variable to increase, only that it causes it to change in the same direction.
Similarly, the "" sign does not indicate a decrease, only that there is a change in the opposite direction.

(a)
(b)
Figure 5: Behavior of (a) central heating system and (b) insulin regulation of blood sugar

Modeling feedback loops: stocks and flows

Understanding the connection between the structure of feedback systems and their dynamic behavior is
essential to understanding and using system dynamics. Causal loop diagrams such as the ones above are
useful in understanding structure, but tell very little about behavior. Modeling a system using the program
STELLA can help your students to gain an understanding of the behavior produced by a given structure.
Before your students are ready to use the computers to begin modeling they need to learn about the
basic building blocks that make up STELLA. These modeling elements are stocks, flows, converters, and
connectors (see Figure 6).

D-4243-3

27

Figure 6: STELLA building blocks

The most important element in STELLA is the stock. A stock is something that accumulates.
Stocks can not change instantly, instead they are raised and lowered by flows. A classic example of a
stock/flow system is a bathtub full of water with a faucet and a drain (see Figure 7). The water is a stock.
It is increased by an inflow of water pouring in from the faucet, and is decreased by an outflow of water
exiting the bathtub through the drain. The stock of water is an accumulation. Faucets and drains can be
turned off and on almost at once, but the stock of water has to change at a rate dictated by its flows. If the
bathtub is frozen in time, the stock of water in the tub is the only variable that can be seen.

Water in Bathtub

Faucet

Drain

Figure 7: Stock/flow diagram of bathtub system

The other two STELLA elements are relatively minor. Converters are used to either hold a
constant value, or to apply an equation and convert a set of inputs into an output. Connectors indicate that
one variable causes another variable to change.
Here is another stock and flow system savings in a bank account. Notice the amount of
savings is the stock, which increases according to the rate set by the inflow of interest.

interest

Figure 8: Stock and flows in a savings account

savings

28

D-4243-3

To show the feedback in the system, we can draw connectors to show cause and effect. For
example, when the savings increases, it causes the interest to increase. The interest is a function of the
savings.

interest

savings

Figure 9: Feedback in a savings account

The savings is not the only thing that the determine the amount of interest, however.
There is a constant interest rate that, with the savings, sets the interest each time unit. This is a circular
converter, and a connector is drawn from it to the interest to show that the rate of interest payment is a
function of interest rate and savings.

interest

savings

interest rate

Figure 10: Complete STELLA model of a savings account

Finally, to show the exact interrelationships between the elements, the equation for the interest is:
interest = interest rate savings

By using stocks, flows, converters, and connectors to pictorially represent the system, your
students are forced to make many of their assumptions explicit. This allows them to more easily test the
validity of their assumptions. By showing that the interest is the rate as which the savings increases,
students can better see how positive feedback produces exponential growth. Finally, by representing the
relationships of our system as stocks increased by flows and flows mathematically related to stocks,
stock/flow diagrams lets us model our system and simulate the behavior on a computer using the STELLA
software.

D-4243-3

29

Savings (dollars)
660

340

1
1

20
0

15

30

45

Time (years)

Figure 11: Graph of saving account showing exponential behavior pattern.

60

30

D-4243-3

Student Worksheet: Modeling an Epidemic

Note: In this section of the paper, positive feedback loops are referred to as reinforcing loops, (R), and
negative feedback loops are referred to as balancing loops, (B). This is just another form of terminology for
the feedback loops.
Purpose:
1) To gain a deeper understanding of the structure and behavior of an epidemic.
2) To be able to use STELLA to build, simulate and test simple models.

0 The Handshaking Game -- What happened?

After your role-playing game with epidemics, you probably discovered that the graph of the number of sick
people over time looked something like Figure 0.1.

This can be broken up into two parts, reflecting

different processes at work during your simulation.

1) Early on, there was a behavior of exponential growth. The number of sick people increased
more and more in the first part of the simulation.
2) This was followed by a behavior of exponential approach. As the number of sick people grew
large, it began to level off, changing less and less, as it approached the total number of
people playing the game.

Sick
People

Total Number of People

Exponential Approach

Exponential Growth
Rounds

Figure 0.1: Behavior pattern of a typical session of the handshaking game

What caused those two distinct things to happen? The rules of the role-playing simulation were
the same from the beginning to the end, yet the graph of the people changed considerably. What would
happen if you changed the rules of the game?

D-4243-3

31

In this lesson, you will create a computer model to explore epidemics and to answer these
questions. The computer will let you represent the game and then change it, allowing you to study the
consequences.

1 Feedback loops
Below are two causal loop diagrams describing how these two feedback loops operate. The arrows show
how a change in the amount of one factor can change the amount of another factor. In the box next to each
arrow put either a S or a O. Mark the arrow with a "S" if the arrow indicates a change in the same
direction. Mark the arrow with an "O" if the arrow indicates a change in the opposite direction.

Sick People

Healthy People

New Infections

New Infections

Figure 1.1: Feedback in an epidemic

The R inside the circular arrow indicates a reinforcing positive feedback loop, and the "B
inside of the other circular loop indicates a balancing negative feedback loop. We call them feedback
loops because changes feed back on themselveswhen a condition or factor changes, the change affects
another factor which causes more changes in the original condition.
Question 1.1
What is different between the two loops shown above? Why is one loop reinforcing
while the other is balancing?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

If we say that at a particular time one feedback loop is stronger or dominant over another we mean
that the system is undergoing behavior associated with the dominant type of feedback at that time. In an

32

D-4243-3

epidemic, one type of feedback is dominant in the early stages of the epidemic and the other type of feedback
is dominant in the later stages.
Question 1.2
What kind of feedback drives an epidemic in its early stages? Look at your causal loop
diagrams and graph above and check to see that your answer makes sense. Explain.
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
Question 1.3
What kind of feedback drives an epidemic in its later stages? How?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

D-4243-3

33

Stocks and flows

STELLA models are made up of two principal elements: stocks and flows. Stocks are things in a system
that accumulate over time. Stocks cannot be directly changed. Instead, every stock has an associated flow
that increases or decreases the stock. For example, water in a bathtub is a stock that is increased by an
inflow of water from the faucet and is decreased by an outflow of water down the drain. A population of
rabbits is a stock that has an inflow of births and an outflow of deaths.
Question 2.1:
Test your knowledge of stocks and flows by labeling each variable as either a stock or a
flow by circling the appropriate answer. What units can each variable be measured in?
variable

type

units

population

stock

flow

factory production

stock

flow

atmospheric pollution

stock

flow

interest on savings

stock

flow

high school students

stock

flow

yearly salary

stock

flow

distance

stock

flow

Question 2.2:
What are the flows associated with the following stocks? What units would you use to
measure the stocks? What units would you use to measure the flows?
stock
money in a bank
computers in a store
nuclear weapons
books in a library
tree forest
heat
distance
velocity

inflow

outflow

units of stock

units of flows

34

D-4243-3

Question 2.3
What is the stock in the reinforcing loop in the epidemic? What is its flow? What are
their units?
______________________________________________________________________
______________________________________________________________________
Question 2.4
What is the stock in the balancing loop in the epidemic? What is its flow? What are
their units?
______________________________________________________________________
______________________________________________________________________

D-4243-3

35

3 Reinforcing feedback
In this lab, you are going to build and simulate a STELLA model of an epidemic. However, y o u w i l l
do this in two parts. Right now, you will create just the early part of an epidemic. To do this you
need to make the following assumptions:
1) Sick People always meet and infect healthy people, causing new infections and hence, more
Sick People.
2) There is an unlimited supply of healthy people. In other, words, there is no
upper limit on the growth of the epidemic.

You are not trying to make an accurate model of the epidemic -- yet. In the next
section you will change assumption two and model the full epidemic from your role-playing simulation.
This model is similar to the first part of the epidemic, but not identical.

Sick People

New Infections

Figure 3.1: Stock/flow causal loop diagram of epidemic positive feedback

36

D-4243-3

Step 1: Start the program STELLA.

If you dont know how to do this, ask a teacher or friend who does. Your screen should look something
like this:

Figure 3.2: Blank STELLA screen

D-4243-3

37

On the top of the screen is a menu of the 4 basic building blocks of STELLA: stocks, flows,
converters, and connectors. To use any of them, just click the mouse button to choose a building block
from the top menu and then click the mouse button again to place the building block it on the main screen.

Figure 3.3: STELLA building blocks

Step 2: Place the stock.

The first thing to do is to place the stock. Choose a stock by clicking on the picture of a stock on the top
of the screen. Place the stock on the screen by clicking the mouse again while the cursor is in the middle
of the screen. Then, while the stock is highlighted, type Sick People to name it.

Step 3: Place the inflow.

Click the mouse button on the flow icon at the left side of the screen. Click and hold the mouse about an
inch to the left of the stock; then hold the mouse button down and drag the flow over the stock until the
stock turns grey. Release the mouse button. Label the flow, New Infections.

Step 4: Place the connector

As the stock of Sick People changes it will affect the flow. You can tell STELLA this by connecting the
stock to the flow with a connector. Choose the connector, move the arrow to the stock and click, then hold
down the mouse and drag the wire to the flow. Dont lift the mouse button until the flow turns
grey.

38

D-4243-3

Your screen should now look roughly similar to Figure 3.4.

Sick People

New Infections

Figure 3.4: Stock/flow diagram

Step 5: Determine the Parameters

This stock/flow structure is the core of your model of an epidemics positive feedback. However,
more information needs to included before you are ready to simulate. In particular, the model needs
parameters, constant converters that serve as inputs to the system. It also needs mathematical descriptions
of the relationship between the number of Sick People and New Infections.
A parameter is an input to a system a value or assumption in a system that we can modify and
see the resulting changes in the behavior. We model parameters as constant values inside of converters.
For example, in a model of a savings account with interest, an important parameter would be the interest
rate (see Figure 3.5). Changing the interest rate will causes changes in how much money is generated
during the simulation.

D-4243-3

39

interest

savings

interest rate

Figure 3.5: Model of savings in a bank. Note that the interest rate is a parameter.

These constants could just be buried in an equation, but it is often a good idea to make them
explicit. Giving the constants explicit names by modeling them as converters reminds you that they exist
and makes them easy to change.
Question 3.1
In the handshaking game, what numerical assumptions were in the rules that we should
make parameters to the model?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Two possible parameters to include are:

1) How many seconds are there in each round?
2) How long does it take to shake someone's hand?

From those two numbers we can calculate:

3) How many handshakes per round?

You may not have initially thought of it, but another important parameter to consider is:
4) The probability of a Sick Person shaking hands with a healthy person during each particular
round.
This last parameter will be particularly important in the second section when we complete the model of the
epidemic.

Step 6: Add these parameters to the model.

40

D-4243-3

Choose the converter from the top menu (it's the circle). Place it near the bottom of the screen and label it,
Seconds per Round (see Figure 3.6). Place a second converter near the bottom and label it "Seconds per
Handshake". Between those converters and the stock/flow combinaton, place a converter labelled
"Handshakes per Round."

Since changes in either one of the first two converters will change the number

of Handshakes per Round, take a connector, and connect Seconds per Round to Handshakes per Round.
With a second connector, connect Seconds per Handshake to Handshakes per Round. Handshakes per
Round will directly affect the infection rate, so put another connector from Handshakes per Round to New
Infections. Make sure the (single) arrows are pointing in the right direction.
Add your last converter near the top of the screen, calling it "Prob of a Sick Person meeting a
Healthy Person." Since this converter will also influence the number of infections, string a connector from
it to New Infections.
Check that your screen matches the diagram below. Do not go on until it looks the same!

Prob of a Sick Person Meeting a Healthy

Sick People

New Infections

Handshakes Per Round

Seconds Per Handshake

Seconds Per Round

Figure 3.6: The plumbing of the epidemic model

Now that you have laid out the plumbing of your model, you need to enter the mathematical
relationships in the model. This includes both the Handshakes per Round, and the New Infections (each
round).
Question 3.2

D-4243-3

41

Write a mathematical equation to determine the number of Handshakes per Round using
Seconds per Handshake and Seconds per Round.
Handshakes per Round = __________________________________________________

Question 3.3
Write a mathematical equation to determine the New Infections (each round) based upon
the number of Handshakes per Round and Prob. of a Sick Person meeting a Healthy
Person (PSPMHP).
New Infections = ________________________________________________________

Step 7: Enter the equation for Handshakes per Round and New Infections
Important! Click the globe in the top left side of the screen. It should now say X 2 .
The globe meant that you were looking at general relationships in the model. The X2 indicates that you are
now going to enter the precise mathematical equations defining those relationships.
Double-click the converter Handshakes per Round. Your screen should look like Figure 3.7. In
the upper left corner are a list of Required Inputs. Each of these variables should be part of your equation.
You can put any of these variables in your equation by clicking on them with the mouse.
Enter the equation you wrote above. Remember, * is the symbol for multiplication and / is
the symbol for division. Click OK when you are done. If you get an error, that means you have entered
the equation wrong try again. If everything goes well, you should find yourself looking at the full
diagram.

42

D-4243-3

Figure 3.7: Entering an equation

Notice that the ? in the variable Handshakes per Round has disappeared. That is because there is
now an equation for that variable. Before you can simulate the model, the ? must be gone from all the
variables.
Double-click the flow New Infections. Enter the equation you wrote above the same way you
entered the previous equation. When you are done, click OK.

The final thing we need to do to make our model complete is to set the initial value of the stock
and the parameters of the model.
Question 3.4:
In the handshaking game you played, what was the initial number of Sick People?
Initial value of Sick People = ______________________________________________

D-4243-3

43

Question 3.5
How long (approximately) did it take to shake a hand? Include the time it took to walk
from one person to another.
Seconds per Handshake = __________________________________________________
Question 3.6
How many seconds did you take per round?
Seconds per Round = _____________________________________________________
Question 3.7
If Sick People never shook the hands of other Sick People, what is the probability a sick
person would meet with a healthy person in a round?
Prob of a Sick Person Meeting a Healthy Person = _____________________________

Step 8: Enter the parameter values.

Double-click on the stock of Sick People. Enter the initial value from your answer above. Click OK
when you are finished. In turn, enter the values you wrote above for Seconds per Handshakes, Seconds per
Round, and Prob of a Sick Person Meeting a Healthy Person.

44

D-4243-3

Step 9: Setting up a graph.

In order to be able to view the behavior produced by our model when it is simulated, you need to set up a
graph. Click the graph symbol at the top of the screen. Then, click again somewhere in an empty space
in your diagram. You should then see an empty graph. Choose Define Graph from the Edit window
and you will see this:

Figure 3.8 An empty graph of Sick People.

To tell the computer to display on a graph the variable Sick People, click Sick People in the table of
allowable inputs, then click > > . Click OK to see your graph again.
Before you go on, you also need to tell the computer how long it should run the simulation.
Choose Time Specs from the Run menu. Set the box next to To: to 5. Important! Also set DT to
.05 (it's needed for this simulation to work properly -- don't worry about what it means). Finally, set the
unit of time by typing Rounds under other. Then click OK.

D-4243-3

45

Figure 3.9: Time Specs screen.

Step 10: Predicting the behavior.

You are now ready to simulate your model. Before you do that, however, draw on the graph below what
you think the behavior of the Sick People in your model will look like. The exact numbers are not
important, just draw the general shape.

Sick
People

time

46

D-4243-3

Step 11: Simulate your model and graph the results.

In the Run menu, choose Run. Draw a copy of the graph on your screen onto the axis below.

Sick
People

time
Question 3.8
Did the graph match your predictions? Why or why not?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Step 12: Save your model

Congratulations! You now have a complete model of the exponential growth aspect of an epidemic. Be
sure to save the file before you leave.

D-4243-3

47

4 Combining feedback loops

In the previous section you modelled the reinforcing loop that drives an epidemic in the early
stages. However, as you found out in the epidemic game, positive feedback is only half of the story.
Your task in this section is to modify the model you built to become a more accurate model of the
handshaking epidemic. In order to do this you will add in a balancing loop that restricts the growth of the
epidemic when the number of sick people rises.
When you made your first model of the epidemic, you made these two assumptions:
1) Sick People always meet and infect Healthy People, who become Sick People.
2) The system has an unlimited supply of Healthy People causing Sick People never to interact
with other Sick People. In other words, there is no upper limit on the growth of the
epidemic.

You will now remove assumption two and replace it with the following new assumption:
C) There is a limited number of people in this simulation. Most people will begin
in an healthy state, and through handshakes, become infected Sick People. As the number of
Healthy People decreases, there will be less meetings between Sick People and Healthy
People, and consequently, less New Infections each month.

Healthy People

Sick People

New Infections

Figure 4.1: Causal stock/flow diagram of epidemic

48

D-4243-3

Load the model you built in the previous section onto your computer.

Your screen should look

roughly like this:

Prob of a Sick Person Meeting a Healthy

Sick People

New Infections

Handshakes Per Round

Seconds Per Handshake

Seconds Per Round

Figure 4.2: Reinforcing feedback epidemic model from section 3

To build a more complete model, the first thing you need to add is a new stock of people
susceptible to the disease in other words the healthy population.
Question 4.1
What should be the outflow from this stock of Healthy People?
______________________________________________________________________

The outflow from the uninfected people is the same as the inflow to Sick People: New Infections. As
people become infected with the disease they leave the stock of uninfected people and enter the stock of Sick
People.

D-4243-3

49

Step 1: Place the stock of Healthy People

Place a new stock on the screen called Healthy People. Connect it to the flow New Infections by holding
down the mouse button, dragging the stock on top of the cloud until the cloud turns grey, and
releasing the button.
Question 4.2
What should the be initial value of Healthy People? How many people began the
epidemic game uninfected?
Initial value of Healthy People = ____________________________________________

Step 2: Set the initial value of Healthy People

Set the initial value of Healthy People to what you wrote above.

Step 3: Determine the behavior of Healthy People in the current model

Make a graph that shows both Healthy People and Sick People. Do this by first double clicking on the
graph picture in your diagram, then choose Define Graph... from the Edit menu. Select Healthy
People and click > > , then click OK. Simulate the model by choosing Run from the Run menu.
Question 4.3:
Draw on the graph below the behavior of both Sick People and Healthy People. Label
the maximum and the minimum values of each graph. How are the graphs related? Has
your graph of Sick People changed from your last simulation?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

50

D-4243-3

Sick
People

time
Healthy
People

time
Question 4.4
Besides the fact that the graphs do not show s-shaped growth, what seems strange about
the graphs? What problems or inaccuracies do you see with the behavior of the Sick
People or the Healthy People?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Although you have included another stock, Healthy People, you have not changed the model to
reflect your new assumptions.

D-4243-3

51

Question 4.5
Under the old assumptions there are three constant parameters that affect the flow New
Infections: Seconds per Handshakes, Seconds per Round, and Prob of a Sick Person
meeting a Healthy Person (each round). Under the new assumptions, one of these
variables will no longer stay unchanged as Healthy People become Sick. Which of these
three variables will be affected?
______________________________________________________________________
______________________________________________________________________
Question 4.6
In a sentence, describe the general relationship between Healthy People and this variable.
Will the relationship be a "S" (change in the same direction) or an "O" (change in the
opposite direction).
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
Question 4.7
What is the relationship between the number of Sick People and this variable?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Step 4: Make the model reflect correctly reflect your causal assumptions
Use connectors to link the stock of Sick People and the stock of Healthy People to the variable you think
should be affected by them.
By now, your epidemic model is almost complete. You have all the elements of your model and
you have the plumbing. All that is needed is to come up with the exact relationship between Healthy
People, Sick People and the variable you chose.
Deciding on an exact relationship is not easy -- in fact it is probably the most difficult part of
model building. Some suggestions on how to make this easier:

52

D-4243-3

1) Remember the epidemic game. Your model should directly reflect your experiences playing
the game.
2)

Think through the situation in terms of actual numbers. Your answer should make sense
both near the beginning and near the end of the simulation.

Question 4.8
If there is 1 Sick Person and 9 Healthy People, what is the probability of a sick person
meeting a healthy person? Show your work.
______________________________________________________________________
______________________________________________________________________
Question 4.9
If there are 2 Sick People and 8 Healthy People, what is the probability of a sick person
meeting a healthy person? Show your work.
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
Question 4.10
If there was 5 Sick People and 5 Healthy People, what value should this variable be?
Show your work.
______________________________________________________________________
______________________________________________________________________
Question 4.11
If there was 9 Sick People and 1 Healthy Person, what value should this variable be?
Show your work.
______________________________________________________________________
______________________________________________________________________

Question 4.12

D-4243-3

53

What is the mathematical equation determining the Probability of a Sick Person Meeting
a Healthy Person based upon the number of Sick People and Healthy People?
Prob of a Sick Person Meeting a Healthy Person = _____________________________
______________________________________________________________________
______________________________________________________________________
Question 4.13
In a sentence, explain how your equation works:
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

54

D-4243-3

Step 5: Enter the equation

Enter in the appropriate variable the equation you wrote above. Your model should now look roughly the
same as Figure 4.3.

Prob of a Sick Person Meeting a Healthy

Healthy People

Sick People

New Infections

Handshakes Per Round

Seconds Per Handshake

Seconds Per Round

Figure 4.3: Complete model of the epidemic

D-4243-3

55

Step 6: Predict the behavior

Your first cut at a complete model is now ready to test. Before you simulate, draw on the two axis
below the dynamic behavior you expect from the Sick People and Healthy People, based on your experience
playing the epidemic game.

Sick
People

time
Healthy
People

time

56

D-4243-3

Step 6: Simulate your model and graph the results:

Sick
People

time
Healthy
People

time
Question 4.14
Why did you get this behavior? What in your equations caused this behavior?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Step 7: If not correct, try again.

If the behavior you got did not match the s-shaped growth you expected, go back and come up with a new
equation and simulate again. Keep trying until your models behavior matches the reference behavior you
expect.

D-4243-3

57

Question 4.15
What was your final equation?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Question 4.16
Is there a balancing loop in your model? Where is it?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

Question 4.17
How does the strength of the balancing loop change from the beginning to the end of
the simulation?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
Question 4.18
How does this change in loop dominance (strength) change the behavior from the
beginning to the end of of the simulation?
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________

58

D-4243-3

Step 8: Model Enhancements

You have completed the major tasks of this worksheet. If there is time, feel free to modify your model.
What happens when you change it? Some suggestions:
Change the length of time each handshake takes.
Add more flows -- make people die or be cured.
Add a stock of immune people with a flow from Sick People to Immune People. See your
teacher for information about delays.
(Difficult) Bring carriers into the model. Add the stock and the appropriate flows.
(Very Difficult) What conditions are needed for a recurring epidemic? Try and create them.